GPT-5 is here – but why does it feel dumber?

In this episode of the KI Kantine, Buschi shares his unfiltered experiences of his first days with GPT-5 – and why the new model initially disappointed him. He discusses how difficult it is to choose the “right” model for a task, what actually happens behind the scenes when using GPT-5, and why OpenAI reactivated GPT-4o after criticism. He also provides practical tips on how to centrally manage and meaningfully compare models via OpenRouter – including his current favorite tool, “Codename Goose”. An episode about expectations, genuine intelligence, and the question: Where is the progress actually?

Anyone working seriously with AI needs to know more than just the name of the latest model. This episode provides practical guidance – and an honest look behind the scenes.

Mentioned tools: Codename Goose OpenRouter

The KI Kantine is a project by Michael Busch – Developer, entrepreneur, and curious canteen philosopher.

**Regularly exploring Artificial Intelligence in the real developer workday during lunch – understandable, practical, and with a touch of skepticism.

New episodes appear regularly – usually right when you’re grabbing a tray.

📬 Questions, feedback, or your own AI experiences? Email me at podcast@ki-kantine.de

All episodes & more: https://ki-kantine.letscast.fm/